Scheduling video programs in near video-on-demand systems

This paper presents an analytical (in contrast to commonly-used simulations) approach to program scheduling in near video-on-demand (NVoD) systems. NVoD servers batch customers' requests by sourcing the same material at certain intervals called phase offsets. The proposed approach to analytical modeling integrates both customers' and service-provider's views to account for the tradeoff between system throughput and customers' partial patience. We first determine the optimal scheduling of movies of different popularities for maximum throughput and the lowest average phase offset. Next, we deal with quasi video-on-demand (QVoD) systems, in which programs are scheduled based on a threshold on the number of pending requests. The throughput is found to be usually greater in QVoD than in NVoD, except for the extreme case of nonstationary request arrivals. This observation is then used to improve throughput without compromising customers' QoS in terms of average phase offset and the corresponding dispersion.